04 Measures PDF

Genetic diversity analysis with
molecular marker data:

Learning module
Measures of genetic
diversity
Copyright: IPGRI and Cornell University, 2003
Measures 1
Contents
f Basic genetic diversity analysis
f Types of variables
f Quantifying genetic diversity:
Measuring intrapopulation genetic diversity
Measuring interpopulation genetic diversity
f Quantifying genetic relationships:

Diversity and differentiation at the nucleotide level
Genetic distance
f Displaying relationships:
Classification or clustering
Ordination
f Appendices
Measures 2
Basic genetic diversity analysis

1. Description of variation within
and between populations,
regions, etc.
M
a
r
k
e
t
d
a
t
a
1
1
0
1
0
1
1
0
0
1
0
0
1
0
Individuals
1
1
0
0
0
1
1
0
1
0
0
1
1
1
0
1
0
0
1
0
1
2. Assessment of relationships
among individuals, populations,
regions, etc.
01
1
1
0
1
0
0
1
01
02
03
04
05
06
02
0.56
03
0.33
0.33
04
0.47
0.26
0.50
05
0.32
0.43
0.37
0.28
06
0.33
0.56
0.56
0.37
0.46
Ind5
3. Expression of relationships
between results obtained from
different sets of characters
Ind3
Ind6
Ind4
Ind2
Ind1
Measures 3
Most of the genetic diversity analysis that we might want to do will involve the
following steps:
1.
Describing the diversity. This may be done within a population or between

populations. It may also extend to larger units such as areas and regions.
2.
Calculating the relationships between the units analysed in step one. This
entails calculating the distances (geometric or genetic) among all pairs of
subjects in the study.
3.
Expressing these relationships with any classification and/or ordination

method at hand. Some of these methods will permit comparing the results
of our molecular study with other types of data (e.g. geographical). In the
slide, Ind1, Ind2, may represent, instead of individuals, populations or
regions.
Types of variable
f Qualitative. These refer to characters or
qualities, and are either binary or categorical:
Binary, taking only two values: present (1) or
absent (0)
Categorical, taking a value among many
possibilities, and are either ordinal or nominal:
Ordinal: categories that have an order
Nominal: categories that are unrelated
f Quantitative. These are numerical and are

either continuous or discrete:
Continuous, taking a value within a given range
Discrete, taking whole or decimal numbers
Measures 4
Examples of qualitative variables:

Binary: e.g. leaf hairiness: present (1), absent (0)
Categorical:
Ordinal: e.g. stalk hairs: rare (1), common (2), abundant (3) or
petiole length: short (1), medium (2), long (3)
Nominal: e.g. petal colour: yellow (1), red (2), white (3), purple (4)
Examples of quantitative variables:
Continuous: e.g. root weight (g); leaf length (cm)
Discrete: e.g. number of stamens: 2, 3, 4, ...
number of fruits: 1, 2, 3,
Categorical variables can be converted to binary variables, but with limitations
because, as we will see later, some similarity coefficients give a weight to the
category of a character and this may bias against other characters being evaluated.
That is, the more categories a variable has, the more weight it has when combined
with other binary or categorical variables with few categories.
An example of converting a categorial variable into a binary one:
Petiole length: short (1), medium (2), long (3)
Short: present (1), absent (0)
Medium: present (1), absent (0)
Long: present (1), absent (0)
Quantitative variables can also be converted to binary variables, for example:
From 0 to 3 fruits: present (1), absent (0)
From 4 to 7 fruits: present (1), absent (0), ...
Quantifying genetic diversity: measuring

intrapopulation genetic diversity
f Based on the number of variants
Polymorphism or rate of polymorphism (Pj)

Proportion of polymorphic loci
Richness of allelic variants (A)
Average number of alleles per locus
f Based on the frequency of variants

Effective number of alleles (Ae)
Average expected heterozygosity (He; Neis genetic
diversity)
Measures 5
Polymorphism or rate of polymorphism (Pj)
A gene is defined as polymorphic if the frequency

of one of its alleles is less than or equal to 0.95 or
0.99
Pj = q 0.95
or
Pj = q 0.99
Measures 6
Where,
Pj = rate of polymorphism
q = allele frequency
This measure provides criteria to demonstrate that a gene is showing
variation.
Its calculation is through direct observation of whether the definition is
fulfilled.
It can be used with codominant markers and, very restrictively, with dominant
markers. This is because the estimate based on dominant markers would be
biased below the real number.
A polymorphic gene is usually one for which the most common allele has a
frequency of less than 0.95. Rare alleles are defined as those with frequencies of
less than 0.005. The limit of allele frequency, which is set at 0.95 (or 0.99) is
arbitrary, its objective being to help identify those genes in which allelic variation is
common.
Reference
CavalliSforza, L.L. and W.F. Bodmer. 1981. Gentica de las Poblaciones
Humanas. Ed. Omega, Barcelona.
Proportion of polymorphic loci
This the number of polymorphic loci divided by the

total number of loci (polymorphic and monomorphic),
that is:
P = npj/ntotal
Measures 7
Where,
P = proportion of polymorphic loci
npj = number of polymorphic loci
ntotal = total number of loci
It expresses the percentage of variable loci in a population.
Its calculation is based on directly counting polymorphic and total loci.
It can be used with codominant markers and, very restrictively, with dominant
markers (see previous slide for explanation).
Richness of allelic variants (A)
f Refers to the number of variants in a sample

f The measure of diversity is (A - 1) variants
because, within a monomorphic population, the
degree of diversity is zero (A - 1 = 0)
Measures 8
For a given gene in a sample, this measure tells how many allelic variants can
be found.
It is sensitive to sample size.
Although the distribution of alleles does not matter, the maximum number of
alleles does.
The measure can only be applied with codominant markers.
Average number of alleles per locus
It is the sum of all detected alleles in all loci, divided

by the total number of loci
K
1/K ni
i 1
Measures 9
Where,
K = the number of loci
ni = the number of alleles detected per locus
This measure provides complementary information to that of polymorphism.
It requires only counting the number of alleles per locus and then calculating
the average.
It is best applied with codominant markers, because dominant markers do not
permit the detection of all alleles.
Effective number of alleles (Ae)

It is the number of alleles that can be present in a
population
Ae = 1/(1 h) = 1/pi2
Measures 10
Where,
pi = frequency of the ith allele in a locus
h = 1 pi2 = heterozygosity in a locus
The measure tells about the number of alleles that would be expected
in a locus in each population.
It is calculated by inverting the measure of homozygosity in a locus.
It can be used with codominant marker data.
Its calculation may be affected by sample size.
This measure of diversity may be informative for establishing collecting
strategies. For example, we estimate it in a given sample. We then verify it
in a different sample or the entire collection. If the figure obtained the
second time is less than the first estimated number, this could mean that
our collecting strategy needs revising.
10
Calculating the Ae: an example

Loci (A, B, C)
Population 2
Population 1
Individual 1
A1 A1
B1 B1
C1 C1
A1 A1
B1 B3
C1 C1
Individual 2
A1 A2
B1 B2
C2 C2
A1 A1
B2 B3
C1 C1
Individual 3
A1 A1
B1 B1
C1 C3
A2 A2
B1 B4
C1 C1
Individual 4
A1 A3
B1 B3
C2 C3
A2 A2
B1 B1
C1 C1
Individual 5
A3 A3
B3 B3
C3 C3
A1 A2
B4 B4
C1 C1
Frequency of allele 1
0.60
0.60
0.30
0.50
0.40
1.00
0.10
0.10
0.30
0.50
0.10
0.00
0.30
0.30
0.40
0.20
0.00
0.30
Heterozygosity (h)
0.54
0.54
0.66
0.50
0.70
0.00
Effective number of alleles
2.17
2.17
2.94
2.00
3.33
1.00
Number of alleles
Measures 11
The table on the slide shows an example for calculating the effective number of
alleles. The two populations each have 5 individuals. For each individual, 3 loci are
analysed, each with a different number of alleles, depending also on the population
(locus A has 3 alleles in population 1 and only 2 alleles in population 2, and so on).
First, allele frequencies are calculated for each locus and each population; then,
heterozygosity in each locus; and finally, the Ae, according to the formula shown in
the previous slide.
11
Average expected heterozygosity (He)

(Neis genetic diversity [D])
f It is the probability that, at a single locus, any
two alleles, chosen at random from the
population, are different to each other
f Three calculations are possible:
A locus with two alleles:
h j = 1 p2 q2
A locus j with i alleles:
hj = 1 pi2
Average for several loci:
H = jLhj/L
f The average He over all loci is an estimate of the

extent of genetic variability in the population
Measures 12
Where,
hj = heterozygosity per locus
p and q = allele frequencies
H = average heterozygosity for several loci
L = total number of loci
The average expected heterozygosity is calculated by substracting from 1 the
expected frequencies of homozygotes in a locus. The operation is repeated
for all loci and the average then performed.
It can be applied to all markers, both codominants and dominants.
The estimated value may be affected by those alleles present at higher
frequencies.
It ranges from 0 to 1.
It is maximized when there are many alleles at equal frequencies.
A minimum of 30 loci in 20 individuals per population should be analysed to
reduce the risk of statistical bias.
12
Calculating diversity with a codominant

molecular marker
Individuals
M
10
11 12
13 14
15 16
17
10
11 12 13 14
15 16
17 18
18
19 20
21
22
23 24
19
21
22
23
25 26
27
28
29
30
27 28 29
30
Gel
Locus A
Locus B
Locus C
Locus D
Locus E
Data scoring
Locus A
Locus B
Locus C
Locus D
Locus E
20
24
25 26
1,1 0,1 1,1 0,1 0,1 0,1 0,1 0,1 0,1 1,0 0,1 0,1 0,1 0,1 0,1 1,1 0,1 0,1 0,1 0,1 0,1 0,1 0,1
1,0 0,1 0,1 1,1 0,1 0,1 0,1
0,1 0,1 0,1 0,1 0,1 1,1 0,1 0,1 0,1 0,1 0,1 0,1 0,1 1,0 1,0 1,0 0,1 0,1 0,1 0,1 0,1 0,1 1,0
0,1 1,1 0,1 1,0 1,0 1,0 1,1
1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0
1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0
1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0
1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0
0,1 1,1 0,1 1,1 0,1 1,0 1,1 1,0 1,1 1,1 1,1 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 0,1 0,1 1,0
1,0 1,0 1,0 1,0 1,1 1,1 0,1 0,1
Measures 13
(continued on next slide)

The top half of this slide shows a drawing of a gel with a size marker on the left (M)
and 30 individuals analysed with a codominant marker, which detected five loci (A,
B, C, D and E). Of these loci, only three are polymorphic (A, B and E).
The bottom half of the slide shows the results of scoring bands per individual and
per locus. Note that, for ease of presentation, no more than two alleles per locus
were depicted. Although the bands belonging to loci C and D were scored (1,0) for
all individuals, scoring was not necessary, because the bands did not yield
information on diversity.
See next slide for calculations.
13
Calculating diversity with a codominant

molecular marker (continued)
Locus
Genotypes
A
A1 A1
A1 A2
A2 A2
Gen. freq. (exp.)
p2
2pq
q2
Individuals (no.)
24
30
Gen. freq. (obs.)
P11 = 0.07
P12 = 0.13
P22 = 0.80
B1 B1
B1 B2
B2 B2
Total
p2
2pq
q2
Genotypes
B
Allele
freq.
Data analysis
Gen. freq. (exp.)
20
30
Gen. freq. (obs.)
P11 = 0.23
P12 = 0.10
P22 = 0.67
E1 E1
E1 E2
E2 E2
Total
p2
2pq
q2
Gen. freq. (exp.)
Hi
Total
Individuals (no.)
Genotypes
hj =
(1 - p2 - q2)
0.13
0.87
0.28
0.72
0.63
0.37
0.23
0.41
E
Individuals (no.)
15
30
Gen. freq. (obs.)
P11 = 0.50
P12 = 0.27
P22 = 0.23
0.46
0.22
Measures 14
1. First, we note that loci A, B and E are polymorphic because they fulfill the
requirement of having allele frequencies below 0.99. Loci C and D are
monomorphic. (exp. = expected value; obs. = observed value.)
2. The proportion of polymorphic loci is P = (3/5) = 0.6 or 60%. That is, the number
of polymorphic loci is divided by the total number of loci analysed.
3. To calculate average heterozygosity (Ho), we:
a. Count how many loci, out of the total, are heterozygous. For instance,
Individual1 has one heterozygous locus (A), Individual2 also has one (E),
Individual27 has 2 heterozygous loci (A and E), ... . In all, 16 individuals
were monomorphic (i.e. they had only one band in each of the five loci),
13 individuals had 1 heterozygous locus and 1 individual had 2
heterozygous loci.
b. Calculate the average observed heterozygosity as:
Ho = [16(0/5) + 13(1/5) + 1(2/5)]/(30) = 0.1
4. The intralocus gene diversity (hj) is calculated for each locus according to the
formula in the top row of the table, giving us locus A = 0.23, locus B = 0.41 and
locus E = 0.46.
5. The average expected gene diversity (Hi) is calculated from the formula in slide
number 12:
Hi = (0.23 + 0.41 + 0.46)/5 = 0.22
14
Calculating diversity with a dominant

molecular marker
Individuals
M
10
11 12
10
11 12 13
13 14
15 16
17
18
19 20
21
19 20
21
22
23 24
25 26
27
28
29
25 26
27 28 29
30
Locus A
Locus B
Locus C
Locus D
Locus E
Data scoring
Locus A
Locus B
Locus C
Locus D
Locus E
14 15
16
17 18
22
23
24
30
Measures 15

The top half of this slide shows a drawing of a gel with a size marker on the left (M)
and 30 individuals analysed with a dominant marker. Five loci are identified (A, B,
C, D and E). Of the five loci detected, three are segregating (A, B and E), while the
other two, C and D, are monomorphic.
The bottom half of the slide shows the results of scoring bands per individual and
per locus. Because we are dealing with a dominant marker, bands are scored as 1
when present or 0 when absent. Scoring the bands for loci C and D can be either
omitted or done as in the slide with 1 for every individual.
The next slide shows the calculations.
15
Calculating diversity with a dominant

molecular marker (continued)
Locus
hj =
(1 - p2 q2)
Allele
freq.
Data analysis
Genotypes
Aa
Aa
aa
Total
Gen. freq. (exp.)
p2
2pq
q2
0.11
0.89
0.18
0.82
0.52
0.48
Hi
Individuals (no.)
24
30
Gen. freq. (obs.)
P1 = 0.20
P2 = 0.80
Bb
bb
Total
2pq
q2
Genotypes
BB
Gen. freq. (exp.)
p2
Individuals (no.)
10
20
30
Gen. freq. (obs.)
P1 = 0.33
P2 = 0.67
Ee
ee
Total
2pq
q2
Genotypes
EE
Gen. freq. (exp.)
p2
Individuals (no.)
23
30
Gen. freq. (obs.)
P1 = 0.77
P2 = 0.23
0.19
0.30
0.50
0.198
Measures 16
1. First, we look at the polymorphism shown by all loci. Loci A, B and E fulfill the
requirement of having allele frequencies below 0.99 and as such can be said to
be polymorphic. Loci C and D are monomorphic. (exp. = expected value; obs. =
observed value.)
2. The proportion of polymorphic loci (P) is P = (3/5) = 0.6 or 60%. The average
heterozygosity (He) cannot be estimated because dominant markers do not
allow discrimination between heterozygous and homozygous individuals.
3. Despite the above (2), the intralocus gene diversity (hj) may be calculated for
each locus using the formula that appears in the top row of the table, column 4,
as follows: locus A = 0.19; locus B = 0.30; and locus E = 0.50.
4. The average gene diversity (Hi) is calculated from the formula in slide number
12:
Hi = (0.19 + 0.30 + 0.50)/5 = 0.198
16
Quantifying genetic diversity: measuring

interpopulation genetic diversity
f Interpopulation differentiation for one locus (gST)
f Interpopulation differentiation for several loci
(GST)
f Populations contribution to total genetic
diversity
f F statistics (Wright)
f Analysis of molecular variance (AMOVA)
Measures 17
Differentiation refers to polymorphic differences between populations at different

levels of structure (populations and individuals).
17
Interpopulation differentiation for one locus

(gST)
gST = 1 (hS/hT)
hS = population diversity
hT = total diversity
Measures 18
Where,
hS = (/( - 1)[1 (1/s)xij2 (ho/2)]
hT = 1 - [(1/s)xij]2 + (hS/s) (ho/2s)
= harmonic average of population sizes
s = number of populations
ho = average observed heterozygosity
xij = estimated frequency of the ith allele in the jth population
The formula in the slide provides a measure of differentiation in terms of
alleles per locus in two or more populations
It ranges from 0 to 1. A negative value may be obtained if an error was made
for sampling or an inappropriate marker system was used.
Because of the complexity of its components, its calculation requires
specialized computer software.
It can be used with codominant markers and restrictedly with dominant
markers. This is because it is a measure of heterozygosity. To have a fair
estimate of the real value, several generations are needed.
18
Calculating the gST

Genotypes
A1 A1
A1 A2
A2 A2
p2 + q2
Pop. 1
20
30
50
0.35
0.65
0.545
Pop. 2
10
20
70
0.20
0.80
0.680
Pop. 3
60
10
30
0.65
0.35
0.545
ho = 1/3(0.3 + 0.2 + 0.1) = 0.20
s=3
(p2 + q2) = 1.77
1/ = 1/n1 + 1/n2 + 1/n3 = 1/100 + 1/100 + 1/100 = 0.03
= 33.33
hs = (33.33/33.33 1)[1 1/3(1.77) (0.20/2(33.33))] = 0.4196

[1/3xij]2 = (1/3(0.35))2 + (1/3(0.65))2 + (1/3(0.20))2 + + (1/3(0.35))2 = 0.1967
hT = 1 0.1967 + [0.4196/(33.33 x 3)] [0.20/(2 x 33.33 x 3)] = 0.8065
gST = 1 (hs/hT) = 1 (0.4196/0.8065) = 0.4797
Measures 19
In this example, we have the number of individuals for each genotype for one locus
(A) in three different populations. Using this number, we want to know the degree of
differentiation in the three populations. In the table, the calculations are followed for
all the necessary elements in the formula shown on the previous slide.
The result (gST = 0.4797) shows significant differentiation between populations in
allele frequencies. We can therefore say that a high percentage of genetic diversity
is distributed among populations.
19
Interpopulation differentiation for several loci

(GST)
GST is the coefficient of gene differentiation
GST = DST/HT
Pop2
HS
HT
DST
DST
Pop1
HS
Pop3
DST
HS
Measures 20
Where,
HT = total genic diversity = HS + DST
HS = intrapopulation genic diversity
DST = interpopulation diversity
(HT/HT) = (HS/HT) + (DST/HT) = 1
GST measures the proportion of gene diversity that is distributed among
populations.
A larger number of loci must be sampled.
Equations are complex and should be calculated with specific computer
software.
For example, assuming that:
HT = 0.263
HS = 0.202
DST = 0.263 0.202 = 0.061
Then, GST = (DST/HT) 100 = (0.061/0.263) 100 = 23.19%. This means that, in
this species, a 23% differentiation among populations exist.
20
Populations contribution to total gene

diversity
The contribution is calculated by removing a
population so that its contribution to the total gene
diversity may be evaluated
CT(K) = (HT HT/K)/HT
CS(K) = (HS HS/K)/HT
CST(K) = (DST DST/K)/HT
Measures 21
Where,
CT(K) = contribution of K to total diversity
CS(K) = contribution of K to intrapopulation diversity
CST(K) = contribution of K to interpopulation diversity
HT = total gene diversity
HS = intrapopulation genic diversity
DST = interpopulation diversity
HT/K = total gene diversity after removing population K
HS/K = intrapopulation gene diversity after removing population K
DST/K = interpopulation gene diversity after removing population K
The measure allows quantifying the variation of total gene diversity when a
population is introduced to or removed from a site (e.g. when introducing a
new variety into a farmers field in an in situ conservation programme).
It also serves to measure the impact of losing a population from a given site in
terms of gene diversity.
It can be used only with codominant markers.
21
F statistics (Wright)
The equation for the genetic structure of populations
is:
(1 - FIT) = (1 FIS)(1 FST)
FIT = 1 (HI/HT)
FIS = 1 (HI/HS)
FST = 1 (HS/HT)
Measures 22
Where,
HT = total gene diversity or expected heterozygosity in the total population
as estimated from the pooled allele frequencies
HI = intrapopulation gene diversity or average observed heterozygosity in
a group of populations
HS = average expected heterozygosity estimated from each
subpopulation
The F statistics allow analysis of structures of subdivided populations. It may also
be used to measure the genetic distance among subpopulations, a concept that is
based on the idea that those subpopulations that are not intermating will have
different allele frequencies to those of the total population.
Genetic distance also provides a way of measuring the probability of encounter
between equal alleles (endogamy). The statistical indexes involved measure:
FIS = the deficiency or excess of average heterozygotes in each
population
FST = the degree of gene differentiation among populations in terms of
allele frequencies
FIT = the deficiency or excess of average heterozygotes in a group of
populations
22
Interpreting FST values

The range of the FST is:
0
(no genetic divergence)
When FST is:

0 to 0.05
0.05 to 0.15
0.15 to 0.25
>0.25
(fixation for alternate

alleles in different
subpopulations)
then the genetic differentiation is:

small
moderate
large
very large
23
Measures 23
Calculating F statistics
Genotype frequency
Pop.
A1 A1
A1 A2
A2 A2
pi
qi
2piqi
0.40
0.30
0.30
0.55
0.45
0.4950
0.3939
0.60
0.20
0.20
0.70
0.30
0.4200
0.5238
HT
2(0.625)(0.375) = 0.4688
po
(0.55 + 0.70)/2 = 0.625
HI
(0.3 + 0.2)/2 = 0.25
qo
(0.45 + 0.30)/2 = 0.375
HS
(0.495 + 0.420)/2 = 0.4575

FIT = 1 (0.25/0.4688) = 0.4667
FIS = 1 (0.25/0.4575) = 0.4536
FST = 1 (0.4575/0.4688) = 0.0241
Measures 24

This slide shows an example of two populations and the analysis of one locus (A).
The allele frequencies are calculated (p and q), as are their averages. The variables
HT, HI and HS are also estimated and used to calculate the F statistics. The analysis
shows low differentiation in allele frequencies among the two populations (FST). We
could conclude that almost all the heterozygote deficit was due to nonrandom
mating within the populations (FIS = 0.4536).
F = fixation index (first column on right of table) is the probability that two alleles
carried by one individual will be the same. It should be calculated only with
codominant markers. If done with dominant markers, a biased estimator may result.
24
Calculating F statistics (continued)

Genotype frequency
Pop.
A1 A1
A1 A2
A2 A2
pi
qi
2piqi
0.25
0.50
0.25
0.50
0.50
0.500
0.0000
0.80
0.10
0.10
0.85
0.15
0.255
0.6078
HT
2(0.675)(0.325) = 0.4388
po
(0.50 + 0.85)/2 = 0.675
HI
(0.5 + 0.1)/2 = 0.30
qo
(0.50 + 0.15)/2 = 0.325
HS
(0.500 + 0.255)/2 = 0.3775

FIT = 1 (0.30/0.4388) = 0.3163
FIS = 1 (0.30/0.3775) = 0.2053
FST = 1 (0.3775/0.4388) = 0.1397
Measures 25
This is another example for which the procedures used in the previous slide were
followed. Differentiation in allele frequencies between the two populations seems
greater (FST = 0.1397), with only a moderate effect of nonrandom mating within the
populations (FIS = 0.2053).
25
Analysis of molecular variance (AMOVA)

f AMOVA is a method for studying molecular
variation within a species
f It is based on a hierarchical or nested model
f It differs from an analysis of variance (ANOVA)
in that:
It may contain different evolutionary assumptions
without modifying the basic structure of the analysis
The driving hypothesis uses permutational methods
that do not require the assumption of a normal
distribution
Measures 26
The different hierarchical levels of gene diversity studied through AMOVA may
include:
1. Continents, which may contain lesser hierarchical levels
2. Geographical regions within a continent
3. Areas within a region in a continent
4. Populations within an area of a region in a continent
5. Individuals within a population in an area of a region in a continent
The mathematical description of the model for situations 3 and 4 can be found in
Appendices 2 and 3, respectively.
The next two slides illustrate how to analyse situation 4.
26
An example of AMOVA
Ind.
Pop. 1
Pop. 2
Pop. 3
X...k
15
21
18
54
A1
A2
A1
A2
A1
A2
X...k2
225
441
324
990
Xi...k2
27
33
28
88
Xijk2
15
21
18
54
X...2
2916
SSa
0.6
MSa
0.3
SSb
11
MSb
0.26190476
SSw
10
MSw
0.22222222
10
11
12
13
14
A1 = 1
Present
A1 = 0
Absent
15
Measures 27
In this table, we show data obtained with 15 individuals from each of three
populations in an analysis with a codominant marker. By means of an analysis of
variance, these data will allow us to calculate the F statistics.
The first step is to convert the bands detected in the gels to binary variables with a
value of either 0 or 1. Then, the sums of presences (1) are calculated so we may
proceed with the sum of squares. Calculations are first done for one population and
continued for the others until we have (X...k). We have i = 15 individuals (effect b), j
= 2 alleles (effect w), k = 3 populations (effect a).
Where,
X...k is the result of summing up all the band presences (1) in the
individuals per population
X...k2 is the result of squaring the number obtained above
Xi...k2 is the result of adding up the squares of the sum of allele
presences in each individual (e.g. Indiv.1 in Pop.1 will be (0 + 0)2 + Indiv.2
in Pop.1 (1 + 1)2 + Indiv. ...)
Xijk2 is the sum of each value squared
SS is the sum of squares for effects a, b and w
An example of calculating SS:
SSa = X...k2/ij X...2/ijk = [990/(15 x 2)] - [2916/(15 x 2 x 3)] = 0.6
MS are the mean squares for effects a, b and w
An example of calculating MS: SSa/dfa = 0.6/2 = 0.3, where dfA refers to
the degrees of freedom for effect a (populations).
27
An example of AMOVA (continued)

SV
df
SS
MS
EMS
w2
+ 2b2 + 2*15a2
Populations
0.6
0.3
Indiv./pop.
42
11
0.26190476
w2 + 2b2
Within indiv.
45
10
0.22222222
w2
Estimates of variances and F statistics

a2
0.0012698
b2
0.0198413
w2
0.2222222
0.24333
FIT
0.086758
FIS
0.0819672
FST
0.0052185
(1 - FIT)
0.91324
(1 - FIS)(1 - FST)
0.91324
Measures 28
Where,
SV = sources of variation
df = degrees of freedom
SS = sum of squares (see previous slide)
MS = mean squares (see previous slide)
2 = total estimated variance
EMS = expected mean squares
w2 = 0.2222222
b2 = (MSb MSw)/2 = (0.26190476 0.22222222)/2 = 0.0198413
a2 = (MSa MSb)/2 15 = (0.3 0.26190476)/2 15 = 0.0012698
2 = w2 + b2 + a2 = 0.24333 (total estimated variance)
Calculating the F statistics has already been explained in slide 22. For this particular
example, they would be as follows:
FIT = (a2 + b2)/2 = (0.0012698 + 0.0198413)/0.24333 = 0.086758
FST = a2/2 = 0.0012698/0.24333 = 0.0052185
FIS = b2/(b2 + w2) = 0.0198413/(0.0198413 + 0.222222) = 0.0819672
The allele frequency differentiation among the three populations is very low
(FST = 0.0052185) and is probably a result of many random matings. More loci need
to be analysed to make a conclusion.
28
Quantifying genetic relationships: diversity

and differentiation at the nucleotide level
f Using sequence data
Intrapopulation nucleotide diversity

Interpopulation nucleotide diversity
f Using restriction data
Variations in banding patterns

Intrapopulation nucleotide diversity
Interpopulation nucleotide diversity
Measures 29
For these calculations, the assumption is made that each nucleotide is a locus.
29
Using sequence data: intrapopulation

nucleotide diversity
It measures the nucleotide diversity among several

sequences in a given region of the genome within a
population (X)
X = n/(n 1)XiXjij
Measures 30
Where,
n = number of sequences under analysis in the individuals of the
population
Xi = estimated frequency of the ith sequence in the population
Xj = estimated frequency of the jth sequence in the population
ij = proportion of different nucleotides between sequences i and j
The measure informs about the degree of nucleotide diversity among several
sequences in a given region of the genome. It is equivalent to the measure of
allelic diversity within a locus.
It ranges from 0 to 1 (0 X 1).
The factors limiting the use of this analytical tool are:
Partial genomic sequences must be available
The equation can only be applied to haploid data
This parameter informs about nucleotide sequences, and the model assumes
haplotypes (haploid genotypes). Even if the study is based on diploid individuals,
sequencing of each copy of the genome is needed.
30
Calculating intrapopulation nucleotide

diversity
Sequence
Freq. Xi
Seq1
TCC T CGAT T ATTC C CAGGGTGC C GATG A AT
5/10 = 0.5
Seq2
TCC A CGAT T ATTC G CAGGGTGC C GATG A AT
2/10 = 0.2
Seq3
TCC A CGAT C ATTC C CAGGGTGC A GATG G AT
1/10 = 0.1
Seq4
TCC G CGAT T ATTC T CAGGGTGC G GATG A AT
2/10 = 0.2
10
1,2 = 2/30, 1,3 = 4/30, 1,4 = 3/30, 2,3 = 4/30, 2,4 = 3/30, 3,4 = 5/30
= 10/(10 1)XiXjij
= (10/9)[0.5 0.2 (2/30) + 0.5 0.1 (4/30) + ... + 1 0.2 (5/30)]
= 0.037
Measures 31
This example has 10 individuals in a population X. For each individual, we analyse

one sequence of 30 nucleotides, and find that individual sequences differ at 5
nucleotide positions (blue). In total, 4 alternative sequences for those 30 nucleotides
are present in the population. In the first column, we note the number of individuals
(n) that have the particular sequence alternatives.
Then, we calculate the number of nucleotide differences in each sequence pair
within the population. For example, 1,2 = 2/30 means that between sequence
1 and 2 there are two nucleotide differences (T vs. A in position 4, and C vs. G in
position 14).
Next, we calculate X for the entire population. The number obtained is 0.037,
or 3.7% nucleotide diversity, based on the sequence analysed in the sample of 10
individuals.
31
Using sequence data: interpopulation

f VXY measures population divergence based on the degree
of sequence variation (1 sequence, 2 populations)
VXY = dXY (X + Y)/2
f VW measures average diversity in a population based on
several sequences
VW = (1/s)X
f Vb measures the total differentiation in several populations
Vb = [1/(s(s 1))]XYVXY
f NST is the relative differentiation
NST = Vb/(Vb + VW)
Measures 32
Where,
VXY = divergence among populations X and Y
X = nucleotide diversity in population X
dXY = the probability that two random nucleotides in populations X and Y
be different
s = number of populations
The measure informs about the level of differentiation among nucleotide
sequences in populations.
It requires sequence data in a sample of individuals for each population.
It needs specific computer software that includes sequence alignment
features.
Some of these are CLUSTAL W, MALIGN and PAUP*.
32
Calculating interpopulation nucleotide

diversity
Nucleotide divergence between X and Y
VXY = dXY (XY)/2 = 0.14 (0.037 + 0.09)/2 = 0.0765
Total differentiation
Vb = [1/(s(s 1))]XYVXY = [1/(2(2 1))]0.0765 = 0.03825
Average diversity in each population
VW = (1/s)X = (0.037 + 0.09) = 0.0635
Relative differentiation
NST = Vb/(Vb + VW) = 0.03825/(0.03825 + 0.0635) = 0.3759
Measures 33
Lets say that we have another population Y, in which the nucleotide diversity for the
same sequence analysed in slide 31 is Y = 0.09.
We also know that the probability that two nucleotides as taken at random are
different in X and Y is 0.14 (dXY).
In this slide, we find the divergence between populations X and Y (VXY), the total
differentiation (Vb), the average diversity in each population (Vw) and the relative
differentiation (NST)..
33
Using restriction data: variations in banding

patterns
Restriction site EcoRI
Fragment 1
Fragment 2
DNA
Indiv. 1
GACTGAATTCCACGGCACTGACGAATTCGAAGTGAATTCTTACTTAAGCTAGCCTGAATTCGATAC
CTGACTTAAGGTGCCGTGACTGCTTAAGCTTCACTTAAGAATGAATTCGATCGGACTTAAGCTATG
DNA
Indiv. 2
GACTGATTTCCACGGCACTGACGAATTCGAAGTGAATTCTTACTTAAGCTAGCCTGAATTCGATAC
CTGACTAAAGGTGCCGTGACTGCTTAAGCTTCACTTAAGAATGAATTCGATCGGACTTAAGCTATG
Fragment 2
No recognition
site for EcoRI
I1
I2
Fragment 2
Fragment 1
Gel
Measures 34
The lack of fragment 1 in Individual2 indicates that it carries a different DNA

sequence at least in that restriction site. A small difference of just two nucleotides,
in the drawing above, is sufficient to make the recognition site for the enzyme to
disappear.
34
Using restriction data: intrapopulation

This measure () is based on the number of

restriction fragments present in two samples
= - (1/r)ln G
(if < 5%)
Measures 35
Where,
r = number of recognition nucleotides of a restriction enzyme
ln G = natural logarithm of the probability that there was no substitution in
the restriction site. Its calculation is:
G = F(3 2G)1/4
F = [Xi(Xin 1)]/[Xi(n 1)]
F = proportion of shared fragments
G = F1/4
n = number of haploid genotypes in the population
Xi = estimated frequency of the ith fragment in the population
The measure estimates the diversity in restriction sites in a sample, because
it relies on the nucleotide sequence of the recognition sites of a given
restriction enzyme.
It informs about the nucleotide substitution in restriction sites. It varies from 0
to 1 (0 X 1).
The equations above can be used with haploid samples, mDNA, cpDNA or
haplotypes.
Reference
Karp, A., P.G. Isaac and D.S. Ingram. 1998. Molecular Tools for Screening
Biodiversity: Plants and Animals. Chapman & Hall, London.
35
Using restriction data: interpopulation

f This measure (VXY) indicates the divergence or

differentiation among populations based on
restriction data
VXY = dXY (X + Y)/2
f This measure is also used with RAPD marker
data
Measures 36
Where,
VXY = divergence or differentiation among populations X and Y
X = restriction diversity in population X
dXY = fragment diversity among two populations = (2/r)ln (GXY)
GXY = FXY(3 2GXY)1/4
G = FXY1/4
FXY = proportion of shared alleles among populations X and Y
= (2XiXXiY)/((XiX + XiY))
XiX = estimated frequency of the i fragment in population X
It estimates diversity in the restriction sites of a sample of two or more
populations. It informs about the nucleotide substitution in the restriction sites.
Computer software such as BIOSYS and GENEPOP are useful. Data
obtained are considered as belonging to haploid organisms.
If used with RAPDs, the value of r is replaced by the primer length (r = 10).
In addition, some assumptions are taken:
The appropriate primers are used
Polymorphism due to insertion or deletion is rare
Similar size fragments in different populations belong to the same locus
Fragments must be identified without error
Software such as RAPDISTANCE and RAPDIS is typically used.
36
Calculating interpopulation nucleotide

diversity
P
o
p
u
l
a
t
i
o
n
Seq.
10
11
12
13
14
15
16
17
18
19
20
Freq. Xi
6/20 = 0.30
A2
5/20 = 0.25
A3
9/20 = 0.45
F = [0.30(0.30 3 1) + 0.25(0.25 3 1) + 0.45(0.45 3 1)] = 0.0325

0.30(3 1) + 0.25(3 1) + 0.45(3 1)
G = (0.0325)1/4 = 0.424591
G = 0.0325[3 2(0.424591)]1/4 = 0.039358
SX = -(1/6) ln (0.039358) = 0.539176
P
o
p
u
l
a
t
i
o
n
Seq.
A1
10
11
12
13
14
15
16
17
18
19
20
Freq. Xi
A1
5/20 = 0.25
A2
13/20 = 0.65
A3
2/20 = 0.10
F = [0.25(0.25 3 1) + 0.65(0.65 3 1) + 0.10(0.10 3 1)] = 0.2425

0.25(3 1) + 0.65(3 1) + 0.10(3 1)
G = (0.2425)1/4 = 0.701743
G = 0.2425[ 3 2(0.701743)]1/4 = 0.272587
SY = -(1/6) ln (0.272587) = 0.216633
Measures 37
In each population, we detected three DNA fragments as a result of a restriction

digest: A1, A2 and A3.
Nucleotide diversity in the regions analysed is larger in population X (X = 0.5392)
than in population Y (Y = 0.2166); thus, X has more gene diversity than Y.
Between populations X and Y, the nucleotide differentiation based on restriction sites
is 0.230766.
2>0.30*0.250.25*0.650.45*0.10@
0.300.25 0.250.65 0.450.10

GqXY
0.141251/4
0.613052
@1/ 4
>
GXY
0.14125 3 20.613052
dXY
2 / 6 ln0.163012
VXY
0.604643 1 0.539176 0.216633

2
VW
1 0.539176 0.216633
2
Vb
1 0.226739
2
NST
0.14125
0.163012
0.604643
0.377905
0.11337
0.11337
0.11337 0.377905
0.230766
37
0.226739
Quantifying genetic relationships: genetic

distance
f The genetic distance between two samples is

described as the proportion of genetic elements
(alleles, genes, gametes, genotypes) that the
two samples do not share
f D = 1 when, and only when, the two samples
have no genetic elements in common
Measures 38
Depending on the similarities of individuals, three representation types of distance

(D) are possible:
D = 1 S, known as linear distance, because it assumes that the relationship
with similarity is linear.
D = (1 S), known as quadratic distance because it assumes that the
relationship with similarity follows a quadratic function, so that, to make it
linear, the square root must be calculated.
D = (1 S2), known as circular distance.
Quadratic
1
0.8
0.8
0.6
0.4
0.6
0.4
0.2
0.2
0
0
0.2
0.4
0.6
Similarity
0.8
Distance
1
0.8
Distan ce
D is ta n c e
Circular
Quadratic
Linear
Linear
0.6
0.4
0.2
0.2
0.4
0.6
Similarity
38
0.8
0
0
0.2
0.4
0.6
Similarity
0.8
Distance models
Calculation of distance, or dissimilarity, follows one
of two possible models:
Equilibrium model
Disequilibrium model
t
d
d1
t+1
t+1
d
d2
Distance remains constant over

time (equilibrium exists between
migration and genetic drift)
Distance changes with time

through migration and
genetic drift
Measures 39
For our purposes, we will use the disequilibrium model. Two alternatives exist:
Geometric distance
Does not take into account evolutionary processes
Based only on allele frequencies
Complex relationship exists between distance and divergence time
Genetic distance
Does not take into account evolutionary processes
Distance increases from the time of separation from an ancestral population
A genetic model of evolution is needed
When should we use geometric or genetic distance?
Geometric distance is used for diversity studies in which comparisons are made
according to morphological or marker data gathered from the operative taxonomic
units (OTUs). OTUs may be individuals, accessions or populations. It can be used
with dominant markers (RAPDs, AFLPs) or codominant markers. Because
evolutionary aspects are not considered, the dendrograms obtained cannot be
interpreted as phylogenetic trees giving information about evolution or divergence
among groups.
In contrast, the genetic distance of any given OTU can be incorporated into phylogeny
studies. The model considers allelic frequencies in OTUs and its mathematical
foundation is different. It can be used with both codominant and dominant markers,
although, with the latter, information is lost because only two alleles can be scored.
Genetic distance with dominant markers, however, requires the examination of two
generations of the same population to measure the segregation of loci (Lynch and
Milligan, 1994).
Reference
Lynch, M. and B.G. Milligan. 1994. Analysis of population genetic structure with RAPD
markers. Mol. Ecol. 3:91-99.
39
Disequilibrium models: geometric distance
f This measures the direct relationship between

the similarity index (s) and distance (D = 1 s)
f Different situations are possible, for example:
Binary variables
Quantitative variables
Mixed types of variables
P number of variables
Measures 40

When analysing molecular data, we deal with binary variables (1,0). These will be
discussed in the following slides.
In Appendix 4, you will find additional information for those cases where you also
must deal with quantitative variables, mixed types of variables and a varied number
of variables. In Appendix 5, an example of calculating geometric distances with
quantitative variables has also been added.
40
Geometric distance (continued)

With binary variables:
Multivariate analysis is used and similarity or

differentiation matrices are formed between the
possible pairs of individuals or operative taxonomic
units (OTUs)
Two similar individuals simultaneously have the

minimum value of distance and the maximum value of
similarity
Distance and similarity are inversely related

Similarity is estimated by the number of coincidence
Measures 41
When using molecular marker data and transforming them to binary data, the
following should be considered:
A speciess ploidy number may mask the presence of allelic series in a locus.
If this happens, genetic diversity will be underestimated when using dominant
markers (presence/absence).
If a marker is codominant, large samples are needed to permit detection of all
possible genotypes, particularly if there are several alleles per locus.
Segregation distortions are common in polyploid species.
Most specialized computer software are designed to analyse diploid species.
Therefore if used with polyploid species, biases may occur on estimating the
various genetic diversity indices.
The reproductive system of certain species has not been studied, so their
inheritance type is not sufficiently known.
The largest coverage (coding and non-coding regions) possible of the
genome of the species under study should be sampled and analysed so that
estimates of genetic diversity are reliable.
41
Calculating allele frequencies for diploids

and tetraploids: dominant marker
Individuals
17 18
10
11
12 13
14 15
16
10
11
12 13
14 15
16 17 18
Locus A
diploid
(2X)
Locus A
tetraploid
(4X)
Binary
matrix
Measures 42
This example of 18 individuals from each of a diploid and a tetraploid species was
analysed with a dominant marker. We are assuming that the banding patterns
obtained are alike. Bands are converted to a binary table in both cases. The
calculations of the frequencies are given in the table below. We can see that, in the
tetraploid, genotype 1, for example, can be either AAAA, AAAa, AAaa or Aaaa;
however the band will only be scored as present (1) the same as it will in the diploid
(AA or Aa).
Locu
s
A
(2X)
A
(4X)
Genotypes
Allele freq.
Diploid
AA, Aa
aa
Total
Gen. freq. (e.)
p2 + 2pq
q2
Indiv. number
14
18
Gen. freq. (o.)
P1 = 0.78
P2 = 0.22
Tetraploid
AAAA, AAAa, AAaa,

Aaaa
Gen. freq. (e.)
0.53
0.47
aaaa
Total
p4 + 4p3q + 6p2q2 +
4pq3
q4
Indiv. number
14
18
Gen. freq. (o.)
P1 = 0.78
P2 = 0.22
0.31
0.69
Allele frequencies should be different in both cases; however, the information loss in
the tetraploid individual is significant. Why? This is because, to estimate the
frequency of the recessive allele a, heterozygotes AAAa, AAaa, Aaaa are not taken
into account. This effect is larger when the ploidy number of the species under
study is unknown.
(e. = expected value; o. = observed value.)
42
Calculating allele frequencies for diploids

and tetraploids: codominant marker
7
Individuals
9 10 11
12 13
14
15
16 17 18
12
14
15
16 17 18
A2 A3
A1 A2
A2 A2
A3 A3
A1 A1
Locus A
diploid
(2X)
A1 A3
A3 A3 A3 A3
A1 A2 A3 A3
A1 A1 A2 A3
M
Diploid
binary
matrix
A1 A2 A2 A3
A1 A1 A1 A1
Locus A
tetraploid
(4X)
7
I
D I
10
I
11
U O
13
(1,0,0) (1,0,1) (0,0,1) (1,0,1) (0,1,1) (1,0,0) (1,0,1) (0,0,1) (0,0,1) (0,1,0) (1,1,0) (0,0,1) (0,0,1) (0,0,1) (0,0,1) (0,1,1) (1,0,1) (0,0,1)
Measures 43
In this example, we have 18 individuals from each of a diploid and tetraploid species
and analysed with a codominant marker. One locus is detected (A) with three alleles
in both situations (A1, A2 and A3).
Calculating the allele frequencies in the diploid individual is not difficult (binary
matrix, bottom of slide). For the tetraploid individual, however, conversion to binary
data is hampered by the fact that individuals with alleles A1 A1 A2 A3 cannot be
distinguished from those with other combinations such as A1 A2 A2 A3 or A1 A2 A3
A3. This situation can only be solved by inference based on estimating the DNA
fragment copy number in the gel.
Genotype
A1 A1
A1
A2
A1
A3
A2
A2
A2
A3
A3
A3
Tota
l
Gen. freq.
(e.)
p2
2pq
2pr
q2
2qr
r2
Indivs. (no.)
18
Gen. freq.
P11 = P12 = P13 = P22 = P23 = P33 =
0.11
0.06 0.22
0.06 0.11 0.44
(o.)
(e. = expected value;
o. = observed
value.)
43
0.25
0.15
0.60
Similarity coefficients for binary variables:

examples
Example of the coefficient value
if
a = 3, b = 1, c = 3, d = 2
Author
Expression
S1
Russel and Rao (1940)
a/n
0.333
S2
Simpson
a/min[(a + b),(a + c)]
0.750
S3
Braun-Blanquet
a/max[(a + b),(a + c)]
0.500
S4
Dice (1945); Nei and Li (1979)
a/[a + (b + c)/2]
0.600
c)]1/2
S5
Ochiai (1957)
a/[(a + b)(a +
S6
Kulczynski 2
(a/2)([1/(a+b)] + [1/(a+c)])
0.612
0.625
S7
Jaccard (1900, 1901, 1908)
a/(a + b + c)
0.429
S8
Sokal and Sneath 5 (1963)
a/[a +2(b + c)]
0.273
S9
Kulczynski 1 (1928)
a/(b + c)
0.750
S10
Sokal and Michener (1958)
(a + d)/n
0.556
S11
Rogers and Tanimoto (1960)
(a + d)/[a + d + 2(b + c)]
0.385
S12
(a + d)/[a + d + (b + c)/2]
0.714
S13
(a + d)/(b + c)
1.250
Measures 44
Indiv.j
Indiv.i
a+
b
c+d
a+c
b+d
Where,
n=a+b+c+d
In the table above we see that:
Indices S1 to S9 give value only to the presence of information
Indices S10 to S13 give value to both presence and absence
Next, we will discuss three indices (in red on top table): Simple Matching (S10),
Jaccard (S7) and Nei-Li (S4).
44
Indexes of geometric distance

Simple Matching Coefficient, or simple
concordance coefficient:
(a + d)/(a + b + c + d)
Jaccard Coefficient:
a/(a + b + c)
Nei-Li Coefficient, or Dice:
2a/(2a + b + c)
Measures 45
These three indices differ in their approach for estimating the number of
coincidences and differences.
The Simple Matching Coefficient considers that absence corresponds to
homozygous loci. It can be used with dominant marker data (RAPD and AFLP),
because absences could correspond to homozygous recessives. An example of
application of the Simple Matching Coefficient for categorical variables is found in
Appendix 6 (click here).
The Jaccard Coefficient only counts bands present for either individual (i or j).
Double absences are treated as missing data. If false-positive or false-negative data
occur, the index estimate tends to be biased. It can be applied with codominant
marker data.
The Nei-Li Coefficient counts the percentage of shared bands among two
individuals and gives more weight to those bands that are present in both. It
considers that absence has less biological significance, and so this coefficient has
complete meaning in terms of DNA similarity. It can be applied with codominant
marker data (RFLP, SSR).
45
Disequilibrium models: genetic distance

f Measures the difference between two genes,
proportional to the time of separation from a
common ancestor
f Several models are possible:
Mutation of infinite alleles
e.g. Neis genetic distance
Stepwise mutation model
e.g. Distance using microsatellites
Mutation in the nucleotide sequence
Measures 46
Mutation of infinite alleles (i.e. isozymes)

Each mutation event gives rise to a new allele.
If 2 genes are the same, no mutation has occurred. If 2 genes are
different, an unknown number of mutations occurred.
The average number of mutations since time t when they diverged from
an ancestor is = 2t, where is the rate of mutation and is multiplied by
2 because we are dealing with 2 independent genes.
The probability that 2 genes are common by descent after time t is
P= e-2t.
Stepwise mutation model (i.e. SSRs)

Mutation is a progressive change so fragments that migrate similar
distances have had few mutations.
In the case of SSRs, mutation is assumed to change the number of
repeats, increasing or decreasing step by step. It can be shown that the
square of the difference in the number of repeats between 2
microsatellites is proportional to the time of divergence from a common
ancestor.
Mutation in the nucleotide sequence

It indicates that the simplest substitution is the mutation of a single base.
The main limitation is the loss of information by not knowing the number
of mutations that could have taken place at one site. To solve that
problem, some methods assume the probability of transition (purine o
purine or pyrimidine o pyrimidine) and transversion (purine o pyrimidine
or pyrimidine o purine).
46
Calculating Neis genetic distance

f The standard Neis genetic distance is:
DXY
ln (IXY)
f It is based on the concept of genetic identity

(IXY):
Ixy
Jxy
(JxJy)
Measures 47

Where,
JX = average homozygosity in population X
JY = average homozygosity in population Y
JXY = average interpopulation homozygosity
Such that,
IXY = 1, if two populations have the same allele frequencies in all sampled
loci
IXY = 0, if two populations do not share the same allele frequencies in all
sampled loci
The value of DXY varies from 0 (where populations have identical allele
frequencies) to infinity (, where populations do not share any alleles).
It assumes that the rate of substitution per locus is equal among all loci and
populations.
This distance estimates the codon differences per locus between two
populations.
47
Calculating Neis genetic distance (continued)

Locus
Alleles
Allele frequencies
Pop.1
Pop.3
Pop.2
A1
0.80
0.74
0.65
A2
0.20
0.26
0.35
Locus heterozygosity
hijk
0.3200
0.3848
0.4550
B1
0.86
0.81
1.00
B2
0.01
0.10
0.00
B3
0.13
0.09
0.00
hijk
0.2434
0.3258
0.0000
D1
0.00
1.00
0.30
D2
1.00
0.00
0.70
hijk
0.0000
0.00
0.4200
Average heterozygosity
Hi
0.0433
0.0547
0.0673
Average homozygosity
Ji
0.9567
0.9453
0.9327
Average interpop. homozygosity
Jii
J1,2 = 0.8733
J1,3 = 0.9346
J2,3 = 0.8986
Genetic identity
Iii
I1,2 = 0.9183
I1,3 = 0.9894
I2,3 = 0.9570
Genetic distance
Dii
D1,2 = 0.0852
D1,3 = 0.0107
D2,3 = 0.0440
Measures 48
We have an example where i = 3 populations, j = 3 polymorphic loci, and there are

10 monomorphic loci with. Moreover there are different numbers (K) of alleles per
locus (e.g. A and D have 2 alleles each and B has 3 alleles).
The table shows the results of calculating the allele frequencies in each population,
as well as the locus heterozygosity. We then calculate the average heterozygosity
and homozygosity (1 - heterozygosity) per population.
Lastly, we calculate inter-population homozygosity and genetic identity, so that we
can estimate Neis genetic distance. Calculations are as follows:
jiijk = 6iij pijk pijk, for example, j1,2jk = homozygosity among populations
1 and 2
j1,2jk = (0.8)(0.74) + (0.2)(0.26) + (0.86)(0.81) + (0.01)(0.10) + (0.13)(0.09) +
(0.0)(1.0) + (1.0)(0.0) + 10 = 11.3533
J1,2 = average interpopulation homozygosity = j1,2jk/13 = 11.3533/13 =
0.8733
I1,2 = genetic identity among populations 1 and 2 = J1,2/(J1J2) =
0.8733/(0.9567 0.9453) = 0.9183
D1,2 = genetic distance among populations 1 and 2 = -ln(I1,2) = -ln(0.9183) =
0.0852
Because we have not yet explained clustering methods, we present the distance
matrix and dendrogram of this example in Appendix 7 (click here).
48
Calculating intrapopulation distance, using

microsatellites
f Intrapopulation distance is the average of the
sum of squares of the differences in the number
of repeats between alleles
Swi
2
2
ii' (aii ai' i' )
2n(2n 1)
f The average intrapopulation distance may be

calculated for all analysed loci (ds)
Sw
(1/ds) jSwj
Measures 49
Where,
aij = size of the allele of the ith copy (i = 1, 2, , 2n) in the jth population
(j = 1, 2, , ds)
n = number of individuals in the sample
Two considerations:
The calculation of distance between two alleles is a transformation of the
number of repeats.
One difficulty in using SSRs to estimate genetic distances is their high rate
of mutation.
49
Calculating interpopulation distance, using

microsatellites
This is the interpopulation component for the
average distance among all allele pairwise
comparisons
SB
2
2
(a
ij a i' j' )
j

j'
i

i'
(2n)2d s(ds 1)
Measures 50
The global distance is the weighted average among the component intra- and
interpopulations
2n - 1
2n(ds 1)
Sw
SB
(2nds 1)
(2nds 1)
These coefficients represent the probability of choosing two different copies of one
locus in the same population and between two populations.
Useful computer software: MICROSAT, BIOSYS, GENEPOP, GDA and POPGENE.
50
Displaying relationships: classification or

clustering
Is the process of grouping (or clustering) objects in
categories or classes based on their common
attributes or relationships. Grouping can be:
Hierarchical:
Essentialist, which tries to unveil the true nature or
form
Cladistic, which is based on genealogy or phylogeny
Evolutionary, which is based on phylogeny and the
quantity of evolutionary changes
Phenetic, which is based on the highest number of
traits of an organism and its life cycle
Nonhierarchical
Overlapping
Measures 51
Hierarchical: a major class that contains minor classes termed branches.
Nonhierarchical: each individual is assigned to a unique group by comparing it

with the initial classes so that its positioning is the most appropriate.
Overlapping: individuals may belong to more than one group.

Classification types refer to procedures of cataloguing objects, organisms, etc.,
and are used in several fields of knowledge. In our case, we use the hierarchical
classification because of the nature of relationships between individuals, that is,
individual, population, accession, variety, etc., are units that cannot be assigned
to two different groups simultaneously.
Reference
Garca, J.A., M.C. Duque, J. Tohme, S. Xu and M. Levy. 1995. SAS for
Classification Analysis; Agrobiotecnology Course, October 1995. Working
document. Centro Internacional de Agricultura Tropical (CIAT), Cali, Colombia.
51
Phenetic classification
f Shows the relationships among samples by using a
similarity index
f A grouping method or distance is selected so that a
tree diagram (dendrogram) or a phenogram (if the
similarity matrix contains phenotypic data) can be
drawn
1
Measures 52
In this example of hierarchical grouping, all characters are given the same weight in
the grouping process.
Total similarity between two groups is the sum of similarity for each character.
It does not consider genealogy.
Phenetic refers to any character used in the classification procedure, whether
morphological, physiological, ecological, molecular or cytological.
52
Clustering methods
f Clustering steps:
Proximity is defined
Each grouping is estimated according to distance
The branches of the dendrogram are built in each
cycle
f Three main methods are:

Simple linkage (or nearest neighbour)
Complete linkage (or farthest neighbour)
Average linkage (or UPGMA)
Measures 53
Other methods for clustering are available such as:

Unweighted pair-group method using the centroid (UPGMC). It is based on
the distance between the mean value for each group.
Weighted pair-group method using the centroid (WPGMC). It takes the OTUs
median value in the groups.
Ward. It works with the sum of the squared distances for pairs of OTUs. It is
also known as the method of minimal variance because, while taking the
squared values, it becomes a very sensitive method (different OTUs will look
more dissimilar and similar OTUs will look even closer). It may be used with
Euclidian distances and molecular data when a high number of DNA bands is
available.
In the next few slides, we discuss in more detail the three methods listed on the
slide above and show an example for each.
53
Simple linkage
f Or nearest neighbour
f It minimizes the inter-group distance by taking
the distance to the neighbour with the highest
similarity
f It works with regular and compact groups, but is
highly influenced by distant individuals. This is
inconvenient when there are different groups
that are not well distributed in space
d(1,2)
Group 1
d(1,2) = minimum distance

between two OTUs
Group 2
54
Measures 54
Simple linkage: an example

(1)
0.30
0.43 0.35
0.28 0.60 0.40
(3)
(2)
C
C
ADB
0.35
AD
0.35
AD
0.30
0.40
(4)
ADB
0.50 0.40 0.30 0.20
0.10
0.0
D
B
C
Measures 55
1. The distance matrix is formed first; then, in a first cycle, the shortest distance is
selected dAD = 0.28.
2. A new matrix is formed by grouping individuals A and D and calculating the
combined distances:
dB(AD) = min (dBA; dBD) = min (0.30; 0.60) = 0.30
dC(AD) = min (dCA; dCD) = min (0.43; 0.40) = 0.40
3. A new matrix is formed by grouping individual B with group (AD) and calculating
the combined distances
dC(ADB) = min (dAC; dCD; dCB) = min (0.43; 0.40; 0.35) = 0.35
4. The dendrogram is drawn.
55
Complete linkage
f Or farthest neighbour
the distance to the individual with minimal
similarity
f It works well with regular and compact groups
but, again, it is influenced by distant individuals
d(1,2)
d(1,2) = major distance

between two OTUs
Group 1
Group 2
56
Measures 56
Complete linkage: an example

(1)
0.30
0.43 0.35
0.28 0.60
0.40
(2)
(3)
AC
DB
0.40
0.43
BD
0.30
0.40
BD
(4)
AC
A
A
0.60 0.50 0.40 0.30 0.20 0.10 0.0
DB
B
D
A
C
Measures 57
1. The distance matrix is formed first; then, in a first cycle, the longest distance is
selected, dBD = 0.60.
2. A new matrix is formed by grouping individuals B and D and calculating the
combined distances:
dA(BD) = max(dBA; dAD) = max(0.30; 0.28) = 0.30
dC(BD) = max(dCB; dCD) = max(0.35; 0.40) = 0.40
3. The new matrix is formed with groups AC and BD, and the combined distances
calculated:
d(AC)(DB) = max (dAD; dAB; dCD; dCB) = max (0.28; 0.30; 0.40; 0.35) = 0.40
4. The dendrogram is drawn.
57
Average linkage
f Or unweighted pair-group method using the
arithmetic average (UPGMA)
the average pairwise distance among all
individuals of the sample
f Most used method
d(1i ,2j) = average distance
between OTUi and OTUj of
groups 1 and 2
Group 1
Group 2
58
Measures 58
Average linkage: an example

(1)
0.30
0.43
0.35
0.28
0.60
0.40
BC
AD
(3)
BC
AD
0.42
(2)
0.35
AD 0.45
AD
0
0.415
(4)
0.5
0.4
0.3
0.2
0.1
0.0
D
B
C
Measures 59
1. The distance matrix is formed first; then, in a first cycle, the shortest distance
is selected, dAD = 0.28
2. Next, a matrix is formed by grouping individual A with D and calculating the
combined distances:
dB(AD) = (dBA + dBD)/2 = (0.30 + 0.60)/2 = 0.45
dC(AD) = (dCA + dCD)/2 = (0.43 + 0.40)/2 = 0.415
3. A new matrix is formed by grouping the individuals with the shortest distance B
with C, and calculating the combined distances:
d(AD) (BC) = (dAB + dAC + dBD + dBC)/4 = (0.30 + 0.43 + 0.60 + 0.35)/4 = 0.42
59
Choosing a clustering method

f First, gather knowledge of the species under
study such as its diversity, reproduction system,
ploidy number and levels of heterozygosity
f Carefully select the genetic characters to
analyse
f Test different clustering methodologies and
assess the level of agreement obtained with
each of them
Measures 60
In addition, it will always be important to combine as much information as

possible. An example may be found in Appendix 8 (click here), in which
both morphological and molecular data are available, and the use of
separate data sets is compared with the use of combined data.
60
Validating the clustering analysis
f External validation
f Internal validation
f Relative validation
f Bootstrapping
Measures 61
External validation:
The matrix distance is compared with other information not used in the
grouping calculations (e.g. genealogy).
Internal validation:
This technique quantifies the distortion due to the grouping method used. It
builds a new similarity or distance matrix, the co-phenetic matrix, directly
from the dendrogram. Validation is calculated by means of a correlation
coefficient between similarity or distance data from the original matrix and
those from the new co-phenetic matrix. Whether the original distances are
maintained is assessed after the grouping exercise (Sokal and Rohlf, 1994).
Relative validation:
Similarity between methods is compared.
Bootstrapping:
This is a re-sampling method by replacement with the same data matrix. It
allows calculation of standard deviations and variances, and is useful for
those situations in which the number of samples or resources (e.g. time,
budget) is limited.
Examples of applying the co-phenetic correlation and bootstrapping methods are
shown next.
Reference
Sokal, R. and J. Rohlf. 1994. Biometry: The Principles and Practice of Statistics
Biological Research (3rd edn.). Freeman & Co, NY.
61
in
Co-phenetic correlation: an example

A
D
Dendrogram
0.30
0.43
0.35
0.28
0.60
0.40
0.60 0.50 0.40 0.30 0.20 0.10 0.0
0
A
Original distance matrix
Co-phenetic correlation =
0.5557
0.43
0.43
0.35
0.28
0.43
0.43
C
0.43 0.35 0.28
Co-phenetic matrix
Measures 62
To construct the co-phenetic matrix, we look at the dendrogram previously built with
the original matrix (this example comes from slide 58). We see that the distance
between D and C in the dendrogram is 0.43, so we fill that cell in the co-phenetic
matrix. Distance between B and C is 0.35, and so on.
Calculations for the co-phenetic correlation are based on the correlation coefficient:
r = (6XiYi - 6Xi6Yi/n)/SXiSYi
Where,
Xi and Yi are the similarity or distance values of the original and cophenetic matrix, respectively
SXi and SYi are the standard deviations for each variable
If the correlation value is high, we can conclude that the dendrogram does indeed
reflect the distances in the original matrix and that therefore there is no distortion
due to the grouping method. In the above example, we obtained a value 0.5557.
This is an average value that could indicate that the dendrogram distances do not
reflect the distance data in the original matrix, and so distortion exists because of
the method used. However, in building the example, we used very few data; nor
were they the real results of an experiment, thus explaining the value obtained.
62
Bootstrapping validation: an example

(1)
A
C
P1
P2
P3
P4
(2)
D
E
Gel
P1
P2
P3
P4
L1
L2
L3
L4
L5
Data matrix
(3)
P1
P2
P3
P1
P2
0.400
P3
0.600
0.400
P4
0.400
0.200
0.400
P4
Similarity matrix
Measures 63

In the gel above (top left corner), we have 4 individuals (Pi) and 5 loci (Lj). We will
suppose we perform the validation in three samples with replacement.
First, we score the marker data in the individuals (data matrix) and then, we
calculate the average similarity (simple matching) and its interval.
63
Bootstrapping validation: an example

(continued)
P1
Average similarity matrix
with standard deviations
P2
P3
P4
P1
P2
0.267 0.115
P3
0.600 0.000
0.400 0.200
P4
0.533 0.115
0.200 0.000
0.400 0.200
Dendrogram before replacement
Dendrogram with replacement

1
0.25
0.44
0.63
0.81
0.11
1.00
0.33
0.56
0.78
1.00
Measures 64
For each individual, the value for each locus is taken, one by one, with replacement
and a sample formed of equal size to the number of loci. The possibility exists that a
locus is selected one or more times. For the example:
M1: L1 L1 L2 L3 L5 (locus L4 was not drawn)
M2: L1 L2 L3 L4 L3
M3: L3 L1 L5 L2 L4
In each sample a similarity matrix is calculated.
Average similarities and their standard deviations are estimated for each individual
pair (1 & 2, 1 & 3, 2 & 3, and so on), and the average similarity matrix is created.
A new dendrogram is built, using the average similarity matrix.
For real situations, more than 100 replacement samples should be created.
64
Displaying relationships: ordination
f Ordination is the arrangement or ordering of

sample units along coordinate systems
f The purpose of ordination, as well as
classification methods, is to interpret patterns in
the composition of samples
Measures 65
Ordination is a multivariate method that complements clustering, and is usually

considered to be an approach that is closer to biological reality.
With ordination methods, we want to represent the relationships of samples in a
simple way by reducing the real situation to a low dimensional space (Gauch,
1982). In doing so, the samples composition is studied as a whole, the statistical
power of the analysis is improved because redundancy is somehow eliminated or
reduced, and the relative importance of different gradients can be determined. Most
of all, we get graphical representations that help us intuitively interpret the
relationships of the different groups of samples.
In principle, ordination is both an exploratory and hypothesis-testing tool. In any
case, the results obtained with ordination methods should always be contrasted with
the available knowledge of the sample being studied and as much as possible with
additional information related to the biological question being addressed in the
research.
Reference
Gauch, H.G., Jr. 1982. Multivariate Analysis and Community Structure.
Cambridge University Press, UK.
65
Useful ordination methods for molecular

marker data
f Principal coordinates analysis (PCoA)

f Nonmetric multidimensional scaling (NMDS)
f Correspondence analysis (CA)
Measures 66
Many ordination techniques existsome are based on distance data or on the

calculations of the so-called Eigenvalues (the sum of all variances for each
character in each component). However, those techniques based on continuous
variables (e.g. principal component analysis or PCA) are not appropriate for use
with marker data. Hence, we discuss only briefly the three listed in the slide above.
More details on the basics of these methods would require a deeper mathematical
understanding of the algorithms involved than what we expect from the average
user of this module. We therefore encourage our readers who want to know more
about these methods to search for ordination methods through the Web. For an
overview, check the site
<http://www.okstate.edu/artsci/botany/ordinate/overview.htm>
Principal coordinates analysis (PCoA) attempts to represent distances between
samples and may accommodate matrices from different dissimilarity measures. It
maximizes the linear correlation between sample distances. When used with
Euclidean distances, the results are identical to PCA.
Nonmetric multidimensional scaling (NMDS) works by maximizing the rank order
correlation and attempting to find the best shape to accommodate the data. This
technique uncovers the basic configuration from the dissimilarity sample matrix.
With NMDS, only the pattern of points is relevant, not the origin, and the
representation may be rotated.
Correspondence analysis (CA) repeats the averages of sample scores and finds
spots where all samples falling in the same spot are as similar as possible and,
simultaneously, samples at different spots are as different as possible.
66
Appendices
Measures 67
2. Analysis of molecular variance: example 1

3. Analysis of molecular variance: example 2
4. Geometric distance
5. Transforming data from quantitative variables: an example
6. Applying the simple matching coefficient for morphological characters
(categorical variables)
7. Calculating Neis genetic distance
8. Morphological and molecular similarities
67
In summary
f The analysis of genetic diversity and structure of
populations involves:
The quantification of diversity and the relationships
within and between populations and/or individuals
The display of relationships
f Molecular data are usually handled as binary

data
f Molecular data can be usefully complemented
with morphological or evaluation data. To do so,
there types of variables can be transformed to
binary variables
68
Measures 68
By now you should know

f The basic steps involved in measuring genetic
diversity
f The major ways to describe genetic diversity within
and among populations
f The correct selection of distance calculation to
assess relationships in the sample of interest
f The differences between alternative clustering
methods
f The options available to validate grouping
f The basic notions underlying the concept of
ordination
f The similarities and differences between clustering
and ordination
69
Measures 69
References
Measures 70
CavalliSforza, L.L. and W.F. Bodmer. 1981. Gentica de las Poblaciones

Humanas. Ed. Omega, Barcelona.
Garca, J.A., M.C. Duque, J. Tohme, S. Xu and M. Levy. 1995. SAS for
Classification Analysis; Agrobiotecnology Course, October 1995. Working
document. Centro Internacional de Agricultura Tropical (CIAT), Cali, Colombia.
Gauch, H.G., Jr. 1982. Multivariate Analysis and Community Structure. Cambridge
University Press, UK.
Karp, A., P.G. Isaac and D.S. Ingram. 1998. Molecular Tools for Screening
Biodiversity: Plants and Animals. Chapman & Hall, London.
Lynch, M. and B.G. Milligan. 1994. Analysis of population genetic structure with
RAPD markers. Mol. Ecol. 3:91-99.
Sokal, R. and J. Rohlf. 1994. Biometry: The Principles and Practice of Statistics in
Biological Research (3rd edn.). Freeman & Co, NY.
70
Next
f Software programs for analysing genetic

diversity
f Glossary
71
Measures 71

04 Measures PDF

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

04 Measures PDF

Transféré par

Droits d'auteur :

Formats disponibles

Genetic diversity analysis with

molecular marker data:

f Quantifying genetic relationships:

Basic genetic diversity analysis

Copyright: IPGRI and Cornell University, 2003

Describing the diversity. This may be done within a population or between

Expressing these relationships with any classification and/or ordination

f Quantitative. These are numerical and are

Examples of qualitative variables:

Quantifying genetic diversity: measuring

Polymorphism or rate of polymorphism (Pj)

f Based on the frequency of variants

Copyright: IPGRI and Cornell University, 2003

Polymorphism or rate of polymorphism (Pj)

A gene is defined as polymorphic if the frequency

Copyright: IPGRI and Cornell University, 2003

Proportion of polymorphic loci

This the number of polymorphic loci divided by the

Copyright: IPGRI and Cornell University, 2003

Richness of allelic variants (A)

f Refers to the number of variants in a sample

Copyright: IPGRI and Cornell University, 2003

It is sensitive to sample size.

The measure can only be applied with codominant markers.

Average number of alleles per locus

It is the sum of all detected alleles in all loci, divided

Copyright: IPGRI and Cornell University, 2003

Effective number of alleles (Ae)

Copyright: IPGRI and Cornell University, 2003

Calculating the Ae: an example

Effective number of alleles

Copyright: IPGRI and Cornell University, 2003

Average expected heterozygosity (He)

A locus j with i alleles:

Average for several loci:

f The average He over all loci is an estimate of the

Calculating diversity with a codominant

1,0 0,1 0,1 1,1 0,1 0,1 0,1

0,1 1,1 0,1 1,0 1,0 1,0 1,1

1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0

1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0

1,0 1,0 1,0 1,0 1,1 1,1 0,1 0,1

Copyright: IPGRI and Cornell University, 2003

(continued on next slide)

Calculating diversity with a codominant

Gen. freq. (exp.)

Gen. freq. (obs.)

Gen. freq. (exp.)

Gen. freq. (obs.)

Gen. freq. (exp.)

Gen. freq. (obs.)

Copyright: IPGRI and Cornell University, 2003

Calculating diversity with a dominant

Copyright: IPGRI and Cornell University, 2003

(continued on next slide)

Calculating diversity with a dominant

Gen. freq. (exp.)

Gen. freq. (obs.)

Gen. freq. (exp.)

Gen. freq. (obs.)

Gen. freq. (exp.)

Gen. freq. (obs.)

Copyright: IPGRI and Cornell University, 2003